Today's Image Capturing Needs: Going beyond Color Management

One of the main concerns in both desktop and pre-press environments is reliable color reproduction. This problem is addressed by the color management systems which are aiming at the production of so-called facsimile color. In order to use color management systems, one should know very well what the color space of the digital representation of the source image is. If this knowledge is not available, the CMS work-flows cannot be followed and more intelligent adaptive color correction techniques are required. Even if the source of the images and the scanning equipment is well-known, people often want to reproduce their originals “better”. In order to produce more appealing images, so-called color editing is required. This kind of editing includes range adjustments, tonal adjustments, saturation enhancement, global and selective color transformations etc. In order to increase productivity, these color corrections should be carried out automatically. The main goal of automatic color correction techniques thus consists of bringing the original images (the source of which might not be known) into a well-known calibrated RGB space such that the reproduction of the images is appealing to the viewer. In order to achieve these goals, the images have to be analyzed and reference points have to be detected. This paper is organized as follows. In the first section, we will introduce a general purpose model for automatic image correction. The general techniques exposed in that section will be illustrated by several case studies in the following sections. In the first case study, we will introduce an automatic tonal correction which has been used in the newspaper business for black and white images. The second case study will briefly describe adaptive techniques which have been used in order to convert negatives to a well-calibrated positive RGB space. The complexity of this technique is relatively low since it only involves a global color correction through the indication of a neutral point (which is equivalent to the specification of a global cast). In the third case study, we will deal with the general problem of automatic image correction of color images from unknown sources. In the last section, we will summarize the obtained results and indicate topics for future research. A Generic Approach to Automatic Image Correction First, we will introduce a formal model that defines a general framework for automatic image correction. Both spatial image corrections (such as unsharp masking, de-screening, noise removal etc.) as well as color corrections (such as tonal corrections, selective color corrections) or a combination (such as the correction of colored patterns) can be considered. Basically, any automatic image correction scheme can be split up in 4 steps. The first step is the most difficult one and deals with the general problem analysis. Following steps deal with image correction as such. Step 1: Problem analysis by studying a test set of images First, the problem needs to be defined. Often, this is done based on a number of images (In)n:1..K to be corrected and the manually corrected images (I’n)n:1..K . A classification of the test set can be realized by studying both the original set and the corrected set. This classification can be formalized by a set of parameters (αn)n:1..L which can be calculated for each source image. The parameter set can be used to derive a set of M corrections (Γn)n:1..M such that for each image In in the test set : o i=1 M Γ i ( In ) ≅ In ' The idea, of course, is that the composition of the transformations should produce good results on arbitrary images as well. The problem is to find a number of characterizing parameters which contain enough information to classify the images and to generate good transformations. There is no general method to describe how to define those parameters. There is a danger in both using too many parameters as well as in using not enough of them. The extreme cases are either using the empty set of parameters or taking all input pixels into consideration. Clearly none of these approaches makes much sense. If only color (non-spatial) corrections are to be applied, using a downsampled version of an input image might turn out to be useful. Other characterizing parameters are f.i.: • one-dimensional histograms • multi-dimensional histograms • frequency analysis (using either classical FFT techniques, windowed FFT techniques or multi-resolution analysis) • filtered versions of a complete image or smaller regions etc. Also parameters that are special to the problems to be addressed can be used. We hereby think, e.g., of parameters such as the hue of the lightest and darkest point, average colors of regions of interest showing skin, snow, grass, sky, etc.